Webinar:
Cyber Security Challenges in GenAI Applications
How GenAI changes the attack surface, and what organisations need to understand when building or deploying AI-enabled applications.
.png)
Webinar details
Generative AI is changing how organisations build products, automate workflows and interact with information. But as GenAI moves from standalone tools into applications, agents, connectors and enterprise systems, it also changes the cyber security risk profile.
The risks are no longer limited to traditional application security. GenAI applications can be influenced through prompts, retrieved documents, emails, web pages, tool calls and generated outputs. This creates new challenges for teams responsible for building, deploying or governing AI-enabled systems.
This webinar will explore the cyber security challenges created by GenAI applications, including prompt injection, excessive permissions in agentic systems and insecure output generation.
The session will also share practical lessons for good practice, helping organisations think more clearly about how to reduce risk when GenAI systems interact with data, tools and users.
Why Attend?
Attend this session to understand how GenAI changes the cyber security conversation.
You’ll leave with a clearer view of the risks that can emerge when models retrieve content, follow instructions, call tools or generate code and outputs that are passed into other systems.
This session is ideal for teams looking to move beyond high-level AI risk awareness and understand the practical security challenges involved in GenAI applications.
Speaker: Dr. Jeremy Bradley, Chief AI Scientist, Cambridge Spark
Agenda:
-
How GenAI changes the attack profile: Exploring why GenAI applications introduce new security considerations beyond traditional software and application security.
-
Prompt injection - direct and indirect attacks: Looking at how instructions hidden in prompts, retrieved documents, emails or web pages can influence model behaviour and potentially hijack intended workflows.
-
Excessive permissions in agentic systems: Discussing the risks that emerge when models can call tools, APIs and connectors with broad or poorly controlled permissions.
-
Insecure output generation: Considering what can go wrong when model outputs are passed unchecked into shells, SQL, browsers, code pipelines or production workflows.
-
Practical lessons for good practice: Sharing key considerations for reducing risk when building, deploying or managing GenAI applications.
This webinar is recorded using WebinarGeek
Watch this webinar if you are:
Watch this webinar if you are:
-
A Cyber Security Leader or Practitioner looking to understand how GenAI applications change the threat landscape.
-
A Technology or Engineering Leader responsible for building, deploying or integrating AI-enabled systems.
-
A Data, AI or Digital Leader exploring how to adopt GenAI while managing security, governance and operational risk.
-
A Product or Transformation Lead working with AI-enabled products, workflows, tools or automation.
-
A Risk, Compliance or Governance Professional involved in setting safe and responsible AI practices across the organisation.
-
A Developer or Technical Practitioner interested in the practical risks of prompt injection, tool use, permissions and insecure output generation.